Recall Picks — sports picks that remember

1 · Simulate a match — 1,000 runs

Offchain model. Edge is flagged only when sim ≥ 70% and bookie ≤ 55%.

Sport

Bookie odds — Team A (decimal, e.g. 1.85)

Team A

Team B

Model win % — Team A (e.g. 71)

Bookie odds — Team B (decimal)

Expected points / goals A (optional)

Expected points / goals B (optional)

2 · Log a pick — before kickoff

The pick is stored with an onchain receipt (Sui checkpoint via Tatum, or a clearly-labelled mock receipt when no key is set).

Team A

Team B

League

Tipster

Your pick (team name)

Event date

Model %

Bookie % (implied, devigged)

Sport

3 · Pick journal

Every pick carries its receipt. Sample / mock = demo history, not a live Mainnet memory.

Loading…

4 · Credibility — who to trust

Recomputed from settled picks. Trust is shrunk toward 50 for small samples, so n=1 can’t score 100.

Loading…

5 · Walrus Memory

Namespace: recall-picks. In mock mode these texts live locally only — they are not on Walrus Mainnet.

Recall query

Loading…

Honest limits: there is no football data on Sui and Tatum has no odds API. The model and bookie odds are offchain; Sui proves the pick existed before kickoff (checkpoint receipt), and Walrus Memory learns from the outcome. Simulation is not a prediction guarantee.